S-PLUS
Program for the Analysis of Accelerated Titration Designs
Accelerated Titration Designs for
Phase I Clinical Trials:
Biometric Research Program staff, in conjunction with
others at the NCI and the FDA, have developed new "accelerated
titration designs" for phase I clinical trials, along with
a dose-toxicity model which can be used to evaluate the toxicity
data. The following gives the manuscript reference and abstract,
along with brief descriptions of two of the designs, and the dose-toxicity
model, which can be used to assign the phase II starting dose, for
the use of investigators conducting phase I studies.
Publication:
Simon, R.M., Freidlin, B., Rubinstein, L.V., Arbuck,
S., Collins, J., Christian, M., Accelerated titration designs for
phase I clinical trials in oncology, Journal of the National
Cancer Institute, vol. 89, no. 15, pp. 1138-47 (1997).
Abstract:
Background: Many patients on phase
I trials are treated at doses below the biologically active level.
Current phase I trials often take a long time to complete and provide
little information about inter-patient variability or cumulative
toxicity.
Purpose: Our objective was
to develop alternative designs for phase I trials that are safe,
reduce the number of patients treated at sub-therapeutic dose levels,
reduce the duration of the trial and provide more information.
Methods: We fit a stochastic
model to data from 20 phase I trials of 9 different drugs. We then
simulated new data from the model with the parameters estimated
from the actual trials and evaluated the performance of alternative
phase I designs on this simulated data. Four designs were evaluated.
Design 1 was a conventional design, similar to the commonly used
modified Fibonacci, using cohorts of 3-6 patients with 40% dose
step increments and no intra-patient dose escalation. Designs 2-4
included only one patient per cohort until one patient experiences
dose-limiting toxicity or two patients experience grade 2 toxicity
(during their first course of treatment for designs 2-3 or during
any course of treatment for design 4). Designs 3 and 4 use 100%
dose steps during this initial accelerated phase. After the initial
accelerated phase, designs 2-4 resort to standard cohorts of 3-6
patients with 40% dose step increments. Designs 2-4 use intra-patient
dose escalation if the worst toxicity is grade 0-1 in the previous
course for that patient.
Results: Only 3 of the actual
trials showed any evidence of cumulative toxicity. The average number
of patients required was reduced from 39.9 for design 1 to 24.4,
20.7 and 21.2 for designs 2, 3 and 4 respectively. The average number
of patients who have grade 0-1 toxicity as their worst toxicity
over 3 cycles of treatment is 23.3 for design 1 but only 7.9, 3.9
and 4.8 for designs 2, 3 and 4 respectively. The average number
of patients with worst toxicity grade 3 increases from 5.5 for design
1 to 6.2, 6.8 and 6.2 for designs 2, 3 and 4 respectively. The average
number of patients with worst toxicity grade 4 increases from 1.9
for design 1 to 3.0, 4.3 and 3.2 for designs 2, 3 and 4 respectively.
Designs with an accelerated first stage without intra-patient dose
escalation also provided substantial reductions in the numbers of
under-treated patients but did not reduce the numbers as much as
the titration designs. The average number of patients with worst
toxicity grade 0-1, 2, 3 and 4 was 10.3, 6.3, 5.2 and 2.2 for design
2 without intra-patient escalation and 7.0, 5.6, 5.2 and 2.8 for
design 4 without intra-patient dose escalation.
Conclusion: Accelerated titration
designs appear to effectively reduce the number of patients under-treated,
speed the completion of phase I trials and provide a substantial
increase in the information obtained. These advantages are achieved
with some increase in the number of patients experiencing grade
3 and 4 toxicities. Accelerated designs without intra-patient dose
escalation also provide major reductions in the number of under-treated
patients although they do not provide each patient with the maximum
opportunity to receive a dose which may provide therapeutic benefit.
We believe that the prospective evaluation of these new designs
is justified.
Brief Description of Accelerated
Titration Design 2:
The dose escalation/de-escalation rules are based
on definitions of dose-limiting toxicity (DLT) and of "moderate"
toxicity. These definitions may be protocol specific. Dose steps
represent 40% increments.
The method used for assignment of dose for the first
course of a new patient depends on whether the patient is entered
during the early accelerated phase or the latter standard phase
of the trial. During the early phase, there is one patient per cohort
and the dose for each patient is one dose level (1.4 dose factor)
higher than that used for the first course of the previous patient.
The accelerated phase ends when one patient experiences DLT during
the first course of treatment or when two patients experience "moderate"
toxicity during the first course of treatment.
When the accelerated phase ends, the dose assignment
for the first course of a new patient is the same as the standard
method. Two additional patients are placed on the dose level at
which the last new patient was treated, and, subsequently, cohorts
of 3-6 patients are started at each dose level. If 0/3 experience
first course DLT, the next cohort starts one dose level (1.4 dose
factor) higher. If 1/3 experience first course DLT, up to 3 more
patients are started at that same dose level. If 2 or more experience
first course DLT, no further patients are started at that dose.
The largest dose level at which less than 2 patients experience
first course DLT is expanded to 6 patients.
The procedure used for intra-patient dose modification
is as follows: If a patient has a worst toxicity of DLT during a
course, the dose is reduced one dose level for the next course.
If the worst toxicity is "moderate" during a course, the
dose remains the same for the next course. If worst toxicity is
less than "moderate" during a course, the dose is increased
one level for the next course.
Brief Description of Accelerated
Titration Design 4:
The dose escalation/de-escalation rules are based
on definitions of DLT and of "moderate" toxicity. These
definitions may be protocol specific. Single dose steps represent
40% increments, but during the early phase of the trial double dose
steps (100% increments) are used.
The method used for assignment of dose for the first
course of a new patient depends on whether the patient is entered
during the early accelerated phase or the latter standard phase
of the trial. During the early phase, there is one patient per cohort
and the dose for each patient is twice the dose used for the first
course of the previous patient. The accelerated phase
ends when one patient experiences DLT during any course of treatment
or when two different patients experience "moderate" toxicity
during any course of treatment. In addition, when the first
instance of moderate toxicity is observed, two additional patients
must have been treated at that dose, or a higher dose, (during any
course) without experiencing moderate or worse toxicity, in order
that the accelerated phase continue. This may require the treatment
of one or two additional patients at that dose.
When the accelerated phase ends, the dose assignment for the first course of a new patient is the same as for the standard method. Ordinarily, this would mean that two additional patients are placed on the dose level at which the last new patient was treated. We do not suggest reverting to the dose at which DLT (or the higher of the two dose levels at which moderate toxicity) was observed, if this dose is lower than that at which the last new patient was treated. On the other hand, in the case of slow accrual, where individual patients may have been dose escalated and experienced DLT beyond the level at which new patients are treated, we suggest that the accelerated phase may be terminated at the dose level at which the DLT (or the higher of the two dose levels at which moderate toxicity) was observed, rather than reverting to the dose level at which the last new patient was treated. In other words, new patients may continue to receive initial doses double that of the previous new patient, until the dose level is reached at which the DLT (or the second moderate toxicity) was observed. In any case, subsequently, cohorts of 3-6 patients are started at each dose level. If 0/3 experience first course DLT, the next cohort starts one dose level (1.4 dose factor) higher. If 1/3 experience first course DLT, up to 3 more patients are started at that same dose level. If 2 or more experience first course DLT, no further patients are started at that dose. The largest dose level at which < 2 patients experience first course DLT is expanded to 6 patients.
The procedure used for intra-patient dose modification
is as follows: If a patient has a worst toxicity of DLT during a
course, the dose is reduced one dose level for the next course.
If the worst toxicity is "moderate" during a course, the
dose remains the same for the next course. If worst toxicity is
less than "moderate" during a course, the dose is increased
for the next course. The increase is 2 dose levels during the early
accelerated phase of the trial, and 1 dose level later.
Choice of the Phase II Starting
Dose:
We do not necessarily recommend as the phase II starting
dose the highest dose level for which less than 2 patients experience
first course DLT. Rather, we suggest that the phase II starting
dose be based on the results of fitting all of the toxicity data,
including subsequent course toxicity, to the following model for
an unobserved continuous variable yij representing toxicity
level,
yij = log(dij + aŚDij)
+ bi + eij
.
In this model, dij represents the dose
administered to patient i during course j; Dij represents
the cumulative dose administered to patient i up to, but not including,
course j; a represents the relative effect
of cumulative prior dose. bi
is normally distributed, with mean 0 and a variance sb2
, representing inter-patient variability. eij
is normally distributed, with mean 0 and a variance se2
, representing intra-patient variability. The unobserved variable
yij translates to the observed discrete toxicity levels
by means of three additional parameters, k1 < k2
< k3 , which divide the line into the 4 regions representing
"minimal" toxicity (usually grade 0-1), "moderate"
toxicity (usually grade 2), DLT, and "unacceptable" toxicity.
Thus, the patient toxicity data determines maximum likelihood estimates
of the parameters a, sb2
, se2 , k1
, k2 , and k3 , which yield estimated probabilities
of seeing moderate toxicity and DLT at the various dose levels,
forming a rational basis for recommending a phase II starting dose.
(Drs. Richard Simon and Larry Rubinstein will be available to collaborate
on analysis of these accelerated titration design phase I trials.)
Download the SPlus Program
for the Analysis of Accelerated Titration Designs (~ 11 KB)
Instructions to run the Downloaded File:
Software requirements: S-PLUS 3.3 or higher (Under
MS Windows). Before running the program, load the functions by typing
source('...../ph1atd'). To run the program, type ph1atd( 'inputfile'
, 'doselist' ) at the S-PLUS prompt, where the "inputfile"
is an ASCII file which contains the patient data in the following
format, one record(line) per course:
Column 1: patient number
Column 2: course number
Column 3: dose
Column 4: toxicity
(Records should be ordered by patient and course number) and "doselist"
is an ASCII file which contains all the doses planned to be used
in the trial listed in a single column in ascending order.
Program Output:
- Maximum likelihood estimates of the model parameters
- Confidence intervals for parameters with non-zero MLE, both
MLE and CI are sent to the screen and file "ph1atd.out"
- Graph of probabilities of grade2+, 3+ and 4+ toxicities (for
the first course) averaged over the population of patients.
(the graph is stored in file "ph1atd1.wmf")
- Graphs of probabilities of grade2+, 3+ and 4+ toxicities (for
the first course) for the patients with beta equal to MLE-SD,
MLE and MLE+SD. (the graphs are stored in file "ph1atd2.wmf")
Note: if the MLE of sigma beta is 0 these
graphs are not produced